Trade-offs in nutrient and sediment losses in tile drainage from no-till versus conventional conservation-till cropping systems

Abstract

Nutrient and soil loss from agricultural areas impairs surface water quality globally. In the Great Lakes region, increases in the frequency and magnitude of harmful and nuisance algal blooms in freshwater lakes have been linked to elevated phosphorus (P) losses from agricultural fields, some of which are transported via tile drainage. This study examined whether concentrations and loads of P fractions, total suspended sediments (TSS), nitrate (NO₃⁻), and ammonium (NH₄⁺) in tile drainage in a clay soil differed between a continuous no-till system combining cover crops and surface broadcast fertilizer (no-till cover crop [NTCC]), and a more conventional tillage system with shallow tillage, fertilizer incorporation and limited use of cover crops (conventional conservation-till, CT). Both sites had modest soil fertility levels. Year-round, high-frequency observations of tile drainage flow and chemistry are described over 4 full water years and related to management practices on the associated fields. There were similar water yields in tile drainage between the two systems; however, losses of TSS, particulate P (PP), and NO₃⁻ were consistently greater from the CT site, which received larger quantities of fertilizer. In contrast, dissolved reactive P (DRP) losses were considerably greater from the NTCC site, offsetting the lower PP losses, such that there was little difference in TP losses between sites. Approximately 60% of the DRP losses from the NTCC site over the 4 years were associated with incidental losses following surface application of fertilizer in fall. This study provides insight into trade-offs in controlling losses of different nutrient fractions using different management systems.

Abbreviations

CC

cover crops

CT

conventional conservation-till

DRP

dissolved reactive phosphorus

FWMC

flow-weighted mean concentration

N

nitrogen

NH₄⁺

ammonium

NO₃⁻

nitrate

NTCC

no-till cover crop

P

phosphorus

PP

particulate phosphorus

TDP

total dissolved phosphorus

TP

total phosphorus

TSS

total suspended sediments

1 INTRODUCTION

Phosphorus (P) losses from agricultural watersheds in the Great Lakes region contribute to the eutrophication of the lakes (GLC, 2023), particularly Lake Erie (International Joint Commission, 2014). In the Mississippi basin, nitrate (NO₃⁻) loads are driving eutrophication in the Gulf of Mexico (Jones et al., 2018). In the temperate Great Lakes, P can impact the magnitude and frequency of harmful algal blooms, but NO₃⁻ can affect their toxicity (Gobler et al., 2016). Ambitious targets have been set to reduce springtime P loads into the Western Lake Erie Basin (Stow et al., 2020). Numerous land-management conservation practices have been recommended and implemented (e.g., GLC, 2023; OMAFRA, 2022). However, recent studies have demonstrated that the application of conservation practices in some landscapes can lead to unintended consequences (Jarvie et al., 2017), or trade-offs (Kleinman et al., 2022) between different contaminants. As such, understanding the magnitude and seasonal variability of nutrient losses (P, NO₃⁻) and soil loss between differing cropping systems is critical to sustaining water quality of downstream ecosystems.

Tile drainage is the dominant runoff pathway from cropped fields in the Great Lakes region. Although surface runoff is the dominant P pathway in undulating fields with coarse to medium grained soils (Plach et al., 2019), tile drains represent the dominant P pathway from the level and near-level clay soils that span across much of the Western Lake Erie Basin (Pease et al., 2018). Nitrate (NO₃⁻) is also an important contaminant in tile drainage (Kladivko et al., 1991; Ren et al., 2022).

No-till management can increase dissolved reactive P (DRP) losses from tile drains at field (Williams et al., 2016), and watershed scales (Jarvie et al., 2017), whereas the effects of no-till on NO₃⁻ losses are less clear (e.g., Tan et al., 1998; Waring et al., 2020). The increase in DRP in tile drain discharge under no-till management is partially due to increased preferential flow (Smith et al., 2015), but also due to increased P stratification in surface soils (Kleinman et al., 2015) as P fertilizers are typically surface applied. In contrast, increased NO₃⁻ losses are primarily due to increased water yields through preferential pathways in no-till soils (Daryanto et al., 2017; Tan et al., 2002). Incorporation of fertilizers may reduce the stratification of P in the soil profile and disrupt macropore networks (Kleinman et al., 2015). However, a trade-off of this practice is that it increases particulate P (PP) and sediment losses in surface runoff (Iavorivska et al., 2021; Kleinman et al., 2022), and possibly in tile drainage. Dissolved reactive P in tile drainage has been emphasized because it is most bioavailable. However, a considerable proportion of PP is also bioavailable (Baker et al., 2014), and PP represents the dominant P fraction lost in some landscapes (e.g., Lam et al., 2016; King et al., 2015; Macrae et al., 2007b).

Increasing awareness of the potential increases in nutrient losses in tile drainage associated with no-till has prompted many farmers to use cover (“catch”) crops to immobilize nitrogen (N) and P to reduce nutrient losses in tile drainage. Cover crops can reduce erosion of surface soil and mitigate P and N losses in surface runoff, but they can also increase DRP losses in surface runoff due to the lysing of cells following freeze-thaw (Liu et al., 2019), although this largely occurs in regions with severe winters and is less prevalent in temperate landscapes (Cober et al., 2018). Studies have shown that cover crops reduce NO₃⁻ losses in tile drainage (e.g., Drury et al., 2014; Hanrahan et al., 2018; Strock et al., 2004), but less is known regarding their impacts on P mobilization in tile drainage. Cober et al. (2019) observed no differences in DRP concentrations in soil water and shallow groundwater beneath plots with and without cover crops. Lozier et al. (2017) demonstrated that P released from cover crops did not increase edge-of-field P losses, likely due to soil contact. These findings suggest it is unlikely that P released from cover crops will translate to increased DRP losses in tile drainage. Waring et al. (2020) found that the effects of no-till and cover crops were comparable in their reduction of NO₃⁻ losses in tile drainage, and the combination of the two practices did not decrease losses any further.

Given the range of competing processes controlling P and N losses in tile drainage, more information is needed to characterize potential water-quality trade-offs in different management systems and for different soils and climate settings. The objective was to compare seasonal and annual water yields and concentrations and losses of DRP, total dissolved P (TDP), total P (TP), total suspended sediments (TSS), NO₃⁻, and NH₄⁺ in tile drain discharge from conservation till (considered conventional for this southwest Ontario region) fields and no-till fields incorporating cover crops.

Core Ideas

• Losses of particulate P decreased, but dissolved P increased with no-till, cover crop.
• Tillage increased erosion of fine-textured soil and associated particulate P into tile drains.
• Greater fertilizer P applied at the tilled site did not translate to greater yields.
• Cover crops and no-till alone is not enough to reduce P in tiles because of potential for incidental losses.
• Elevated P losses in tile drainage can still occur from sites without high P soils.

2 STUDY SITES AND METHODS

2.1 Study sites

This study was conducted on two farms in the Chatham-Kent region of Ontario, Canada, on the northern boundary of Lake Erie (42.2410° N, 82.2330° W). Soil in the region and on both farms is Brookston Clay (mesic Typic Argiaquoll). The regional topography ranges from level to gently undulating, and there is extensive use of tile drainage.

The region experiences a warm summer continental climate with mean annual temperatures of 9.8°C and 882 mm mean annual precipitation (10% as snow) (ECCC, 2022). Precipitation is evenly distributed throughout the year. The region is the coolest in winter (January/February) when mean daily temperatures are approximately −3°C; however, day-to-day weather can vary substantially, and extreme daily temperatures have reached maximums of 19°C and minimums of −27°C during January/February (ECCC, 2022). Consequently, there are frequent mid-winter thaw cycles and therefore snow cover is intermittent. In contrast, summer temperatures (June–August) average 21°C, with mean daily maximums of 26°C and daily minimums of 16°C (ECCC, 2022). Spring thunderstorms are common. Highly variable weather conditions coupled with the flat, clay-rich landscape pose significant challenges, forcing farmers to adapt cropping systems in response to conditions.

The two study farms are adjacent, and farmers strive to implement a corn-soy-soy-winter wheat crop rotation (Table S1). However, seasonal precipitation patterns led to the two farms being out of phase in their crop rotations for a significant portion of the study with primarily soybean crops (Tables S2 and S3). Nevertheless, one full 4-year cropping cycle and 4 water years were captured. No manure is used at either site, and fertilizers are surface broadcast. One field (conventional conservation-till, CT) incorporates fertilizer via disk tillage (∼5 cm depth) within 48 h of application (conventional management for the region), and the second site (continuous no-till cover crop system, NTCC) has been under no-till since monitoring operations began in 2017 (last tilled in 2015 to level the field). At the CT site, tillage is also used as a method to manage residues. In contrast, at the NTCC site, cover crops are planted after each annual crop to have a living cover every non-growing season. Phosphorus applications at the NTCC site are surface broadcast onto a living cover crop, and both soybeans and corn are seeded into cover crops at planting.

There is no clear difference in soil organic matter, pH, potassium, magnesium, or nitrogen concentrations in the surface soils of the two sites (Table S4). In general, the Olsen-P content was similar between the two sites, although there is greater P stratification at the NTCC site than at the CT site (Table S4). The NTCC site only received fertilizer application twice during the 4 water years (October 2018 and July 2019). It should be noted, however, that fertilizer was applied at the NTCC site in 2016 (May), prior to the initiation of this study and again in October 2021. In contrast, fertilizer was applied four times over the study period at the CT site (annually each fall). Fertilizer was applied at the CT site prior to planned winter wheat or corn; however, precipitation challenges led the farmer to pivot to planting soybeans in multiple years (Table S2). In fact, both farms planted winter wheat in fall 2018 but terminated crops due to poor establishment and re-planted in spring 2019 (Tables S2–S3).

2.2 Field methods

Hydrometric information (precipitation [snow and rain], air and soil temperature, soil moisture) was determined using an on-site meteorological station (ADCON), recording data at 15-min intervals. Two tile drain “mains” were installed in each field (2015 at NTCC and 2016 at CT) and connected to existing subsurface tile laterals in each field to permit the instrumentation of two ∼4 ha study plots per field. In each of the tile drain mains (2 per field), flow was measured using Flo-tote 3 sensors (depth, velocity, flow direction) and recorded at 15-min intervals on an FL-900 data logger (Hach Ltd.). Pressure transducers (U20; Onset Ltd.) were also placed in tile drains to capture water levels during periods in which the Flo-tote sensors failed. Water samples were taken at high-frequency (2–12-h intervals) during storm and thaw events using Teledyne ISCO 6700 autosamplers from each of the four monitored mains. Water samples were retrieved within 24–48 h of collection, and immediately processed in the laboratory. For each sample, a ∼50 mL aliquot was passed through a 0.45-μm pore size cellulose acetate filter. An unfiltered 100-mL aliquot was acidified to a final concentration of 0.2% (v/v) H₂SO₄ and stored in the dark. Unfiltered samples were digested using acid persulfate in an autoclave and subsequently analyzed (EPA/600/R-93/100, Method 365.1). Filtered water samples and digested unfiltered water samples were analyzed colorimetrically (Lachat Instruments 2017–2018 at the University of Guelph; 2019–2021 Bran Luebbe AAIII at the University of Waterloo) for DRP (filtered sample) and total P (TP) (unfiltered sample). TDP, NO₃⁻, and NH₄⁺ were analyzed on filtered water samples collected from 2019 to 2021. Filtered water samples were digested by UV and acid persulfate using an inline UV-digester, and subsequently analyzed colorimetrically for TDP (Bran Luebbe AA3, Seal Analytical, Method No. G-092-93- Rev1). Soluble nitrogen (N-NO₃⁻ and N-NH₄⁺) was also analyzed colorimetrically (Bran Luebbe AA3, Seal Analytical, Methods G-102-93 [NH₄⁺], G-109-94 [NO₃⁻ + NO₂⁻]). Unfiltered samples were analyzed for TSS, where ∼100 mL of sample was passed through a 0.45-μm pore size filter, and the dry mass of the material on the filter was obtained gravimetrically. Only a subset of samples analyzed for P fractions were also analyzed for TSS, NO₃⁻, and NH₄⁺.

Soil samples (0–2 and 0–15 cm) were collected by Agriculture & Agri-Food Canada (AAFC) using standard techniques in both fields for fertility analyses twice over the study period (Table S4) and analyzed at A&L Labs (London, Ontario, Canada). Additional samples were collected at a subset of locations in the fields in 2017 and 2020 by OMAFRA (not shown) and found to be comparable to the AAFC data; however, due to the smaller number of data points and variability in sampling, only the AAFC data have been reported here (Table S4). Textural analysis was also determined in 2018 using the pipette method.

2.3 Data processing and correction, and statistical analyses

Although backflow has been observed at other sites (e.g., Marshall, 2020; Van Esbroeck et al., 2016), necessitating the measurement of both depth and velocity of tile drain flow, it did not appear to be a significant phenomenon in the current fields under most conditions, and consequently depth-discharge relationships were developed for each of the four tile drains, permitting gap filling for periods in which the depth-velocity sensors failed, and only depth data were available. During periods when both sensors failed, tile drainage water yields were estimated for an individual tile using relationships between that tile and the other three tile drains in the study, where the average of predictions for the unknown tile from each of the other three tiles was used. The modelled data were not used in any statistical analyses. The two plots for each field were averaged together to produce estimates of tile drain discharge from each farm.

Events were delineated using HYDRUN (Tang & Carey, 2017), where events commenced when tile discharge rose above baseflow and ended when discharge reached the next local minimum. Flow weighted mean concentrations (FWMCs) were determined for individual events for each of the four tile drains using continuous flow data and discrete measures of DRP, TP, and TSS (after Williams et al., 2016). For events not captured by autosamplers, loads were estimated for each field (using the two plots on each field) and then interpolated using flow-load relationships (see Lam et al., 2016; Macrae et al., 2007a). During baseflow periods, P and TSS concentrations are typically very low (∼0.01 mg/L [DRP] and ∼0.01–0.02 mg/L [TP] based on previous data collection at the site) and consequently contribute very little to edge-of-field losses. As such, event-related losses are focused on in this study. Seasonal and annual nutrient and TSS losses are provided for each farm (the two plots on each field are averaged together).

Nutrient and TSS concentrations and losses did not meet the assumption of a normal distribution, so non-parametric methods were used for statistical analyses. Sheirer–Ray–Hare tests using the effects of Farm and Season as factors were used to examine differences in FWMC during all events captured by autosamplers. One-way Kruskal–Wallis tests with post hoc analyses (Dunn's tests) were used to examine seasonal differences in event-related flow or FWMC for each farm.

3 RESULTS AND DISCUSSION

3.1 Annual and seasonal water and nutrient yields and speciation in tile drainage

Annual tile drainage over the 4 water years averaged 202 mm ± 48 mm at the CT site and 183 mm ± 23 mm at the NTCC site (Figure 1), which is comparable in magnitude to what has been observed at other sites in similar landscapes in the Lake Erie watershed (e.g., Pease et al., 2018; Plach et al., 2019; Wang et al., 2016; Zhang, Tan, Wang, et al., 2017; Zhang, Tan, Zheng, et al., 2017). Event-based and seasonal water yields did not differ across sites (Dunn's test, p > .05). Although surface runoff was not measured in the current study, a multi-year study at a nearby site found tile drainage to represent ∼80% of edge-of-field water years (Plach et al., 2019). A wide range of conditions were experienced over the study period, which is typical for the southwestern Ontario region in which day-to-day weather is highly variable. Overall, conditions were wetter and cooler than normal in 2017–2018 (1028 mm, 9.6 C) and 2018–2019 (978 mm, 9.3 C), but typical in the 2019–2020 and 2020–2021 water years. These differences in precipitation and temperature led to slightly greater water yields from tile drains in the wetter years (∼250 mm annual water yields in 2017–2018, 2018–2019, and ∼195 mm water yields in 2019–2020 and 2020–2021). In all study years, event-related runoff losses (Figure 1) accounted for more than 90% of annual subsurface runoff losses.

FIGURE 1

Four-year seasonal mean losses of tile drainage water (mm), dissolved reactive phosphorus (DRP) (g ha⁻¹), total P (TP) (g ha⁻¹), total suspended sediments (TSS) (kg ha⁻¹), and NO₃-N (kg ha⁻¹) occurring during major events for the seasons of each water year for the two farms (spring = April, May, June; summer = July, August, September; fall = October, November, December; winter = January, February, March). Data shown in the figure include both observational and interpolated data to generate budgets. Values presented for each farm are the average of the two monitored plots on each farm (conventional conservation-till [CT], no-till cover crop [NTCC]).

Although hydrological losses occurred throughout the year, the majority of annual runoff consistently occurred during the non-growing season in each year (77%–85% in a given water year, Figure 1). This was consistent across tile drain plots and in both wetter and drier years. Of the growing season losses, the majority of these were associated with the early growing season (May–June events), and minimal water yields occurred in summer (July–September) in most years (Figure 1). This is despite the fact that rainfall occurred during those months. It is likely that much of the rainfall that occurred was either lost via evapotranspiration or stored in the soil unsaturated zone. Preferential connectivity in clay soils, particularly those under no-till, can enhance infiltration and water transfer into tile drains (Kleinman et al., 2015); however, this does not appear to have led to appreciable differences in water yields between no-till and conventionally managed soils in this study (Figures 1 and 2). Very few summer rainfall events generated flow (Figure 1), with the exception of 2021, presumably due to the increased hydrologic storage potential in soils due to summer crop growth. The greater summer runoff and associated nutrient and TSS losses in 2021 were driven by larger than normal precipitation.

FIGURE 2

(a) Annual precipitation (lines) and tile discharge (blue bars) for the conventional conservation-till (CT) and no-till cover crop (NTCC) sites and (b) phosphorus fractions lost annually from the two farms for each of the water years studied. Dissolved reactive phosphorus (DRP) is shown in red, dissolved unreactive phosphorus (DUP), calculated as the difference between total dissolved phosphorus (TDP) and DRP, is shown in green, and particulate phosphorus (PP), calculated as the difference between total P (TP) and TDP, is shown in blue. An average of the two plots on each of the two farms (CT, NTCC) is shown in the figure.

There were subtle differences in tile drainage water yields between the treatments, but such differences were not consistent across all years, and over the duration of the entire study period, total tile drainage water yields did not significantly differ (p > 0.05) between the two fields. Total tile drainage water yields were greater from tiles in the CT field compared to the NTCC site in the 2018–2019 (60% more) and 2019–2020 (25% more) water years, but smaller in the 2017–2018 (10% less) and 2020–2021 water year (30% less). Of these study years, cover crops were planted in the 2017–2018, 2018–2019, and 2020–2021 water years.

Annual DRP losses (mean ± standard deviation) were nearly two times greater from the NTCC site (0.93 ± 1.0 kg/ha) than the CT site (0.52 ± 0.3 kg/ha) in most years (Figure 1), which is consistent with what has been shown in other studies that have reported greater P losses from NT than CT (e.g., Williams et al., 2016). In contrast, annual TSS and NO₃⁻ losses were approximately twice as large at the CT site (TSS: 1171 ± 425 kg/ha; NO₃-N: 11.2 ± 7.9 kg/ha) than at the NTCC site (TSS: 543 ± 171 kg/ha; NO₃-N: 7.8 ± 5.2 kg/ha). Total P losses were ∼30% greater from the CT site (2.4 ± 0.6 TP kg/ha) than the NTCC site (1.8 ± 1.1 kg/ha) in most years, with the exception of the 2018–2019 water year when P was surface applied prior to planting no-till winter wheat in the fall at the NTCC site (Table S3). The differences in cumulative nutrient losses between the sites were driven by a small number of events.

Total P losses were largely as PP from both farms (Figure 2), with the exception of the NTCC farm during the 2018–2019 water year when fertilizers were surface applied to the NTCC site (Table S3). The greater TP losses from the CT site were primarily due to elevated PP losses in the tile drainage from that site (Figure 2). Total dissolved P (TDP) and DRP concentrations were positively correlated (R² = 0.996; p < 0.001 across both sites), where a median of 81% of TDP was as DRP (Figure 2). Most (98%) of the inorganic N losses (NO₃⁻+NH₄⁺) were as NO₃-N.

3.2 Drivers of temporal variability in nutrient concentrations and yields in tile drainage

There was considerable variability in FWMC of nutrients across the events within and among seasons (Figure 3), and clear patterns were not apparent. For all events captured by autosamplers, Kruskal–Wallis tests found significantly greater TP (p = 0.006) and TSS (p < 0.001) concentrations at the CT site but did not find significant differences in DRP, NO₃⁻, or NH₄⁺ concentrations between the two farms (all events pooled). A significant interaction was found between Farm and Season for DRP concentrations (p < 0.05), but no significant interactions were found for any of the other nutrients studied. Significant effects of season were found for TP (p < 0.001), TSS (p < 0.001), and NO₃⁻ (p = 0.024) but not for NH₄⁺ (p > 0.05) (data from both farms pooled) or for DRP on each farm (p > 0.05). Although there were differences among some seasons for specific nutrients (e.g., the summer period was lower in TP and TSS than other seasons at the CT site, but not the NTCC site; NO₃⁻ was greater in spring than other seasons at the CT site), clear seasonal differences were otherwise absent. Significant differences between farms for individual seasons were only found for TSS in the spring, winter, and fall, when TSS FWMC from the CT were greater than from the NTCC site.

FIGURE 3

Water yields (Flow) and flow-weighted mean concentrations of dissolved reactive phosphorus (DRP), total P (TP), total suspended sediments (TSS), NO₃⁻-N, and NH₄⁺-N during events for each season for all years (spring = April, May, June; summer = July, August, September; fall = October, November, December; winter = January, February, March) on each of the study farms. Only observed data are shown. Boxplots show the median (center black line), the 25^th, and 75th percentiles (bottom and top of the boxes, respectively), and error bars show the 10th and 90th percentiles. Outliers are shown with circles. Data shown are from both monitoring plots (i.e., individual tiles rather than averages of the plots) on each of the two farms (conventional conservation-till [CT], no-till cover crop [NTCC]) are shown. Some extremes were observed that have been removed from the figures to ease readability (DRP concentrations of 3000–5000 μg/L were observed from the NTCC plots in fall 2018). Differences are significant among plots and seasons as indicated by labeled letter differences.

Previous research on tile drains in clay soils has reported elevated P concentrations in tile drains when surface runoff is rapidly routed into tile drains through preferential pathways (Smith et al., 2015). However, such pathways are less active during the non-growing season when soils are wetter and more swollen (Macrae et al., 2019; Grant et al., 2019a, 2019b). While elevated P concentrations associated with smaller water yields (and presumably preferential flow) at no-till sites were observed at other sites during summer events by Kokulan et al. (2021) and King et al. (2017), this was not observed in the present study.

In general, nutrient and TSS losses reflected discharge losses as nutrient and sediment loads are positively related to discharge (Lam et al., 2016); however, some notable exceptions occurred in the current study that were related to management practices on fields (Figure 4). The elevated DRP losses in the 2018–2019 year at the NTCC site (Figures 1 and 2) coincided with a broadcast fertilizer application in the fall of 2018. This influenced overall DRP losses from that site over the entire study period and demonstrates the importance of incidental losses where fertilizer application coincides with a hydrologically active non-growing season. Examination of differences in FWMC for individual events between the two sites during periods that followed fertilizer application (first event within 1–2 months of application) and those that did not demonstrate differences in DRP and TP concentrations (Figure 4) between the two sites. DRP concentrations were most often greater from the NTCC site (all events, Figure 4), but this was especially apparent for events that followed P application at both sites (p = 0.006, Figure 4) or when fertilizer was applied only at the NTCC site but not the CT site (Figure 4) when compared to events for which no fertilizer was applied. When fertilizer was applied at the CT site but not the NTCC site, FWMC of DRP were similar between the sites (p = 0.78). In contrast, TP concentrations were greater at the CT site for most events (all events, Figure 4). This was true for events that did not follow fertilizer application at the NTCC site (none, Figure 4), or when fertilizer was applied to the CT site but not the NTCC site; however, when fertilizer was applied at the NTCC site or both sites, TP concentrations were greater at the NTCC site. Differences in TP concentrations between sites were not significant between events with no fertilizer application and events following fertilizer application at the CT site (p = 0.72) but were significant for events following fertilizer application at both the CT and NTCC sites (p = 0.006). Too few events occurred following P application at the NTCC site alone for statistical analyses. PP losses (Figure 3) and TSS concentrations (Figures 2 and 4) were greater at the CT site than the NTCC site over the study period, and this pattern was consistent across the study years. However, there was no difference in TSS between events that followed tillage and those that occurred during periods without recent tillage (Figure 4, p > 0.05). Unfortunately, fewer events immediately following fertilizer application were analyzed for N fractions, or events did not occur within 1–2 months of N application, and consequently the impacts of N application on N concentrations (incidental losses) cannot be explored here.

FIGURE 4

Differences in flow-weighted mean concentrations between the two farms (flow weighted mean concentrations [FWMC] at conventional conservation-till [CT]-FWMC at no-till cover crop [NTCC]) for each of the sampled events for (a) dissolved reactive phosphorus (DRP), (b) total P (TP), and (c) total suspended sediments (TSS). Box-whisker plots show the differences in FWMC across the population of events sampled (median [center black line], the 25th and 75th percentiles [bottom and top of the boxes, respectively], and error bars show the 10th and 90th percentiles), as well as individual events (symbols) which are colored by the discharge (mm) for each event. Distributions are shown for all events (All), as well as those that did not occur within 1–2 months following fertilizer application or 3 months following tillage (none). Events that were the first or second event following P application (but within 1–2 months of P application or 3 months of tillage) at either the CT site, the NTCC site or both CT and NTCC sites are also shown. Nitrogen concentrations are not shown as fewer events were sampled, and those immediately following fertilizer application were not always captured. Values > 0 indicate greater concentrations at the CT site, whereas values < 0 indicate greater concentrations at the NTCC site.

In the current study, the NTCC farmer applied P on the surface prior to planting winter wheat, with the goal of having the emerging crop reduce P losses in runoff. Any winter wheat emergence, however, did not prevent the occurrence of incidental losses during the first few events that followed the P application. Although winter cover crops can release P following freeze-thaw cycles (Liu et al., 2019), this does not appear to have been significant in the current study as losses and concentrations of P were not greater in winter or spring than at other times of year. Given the incidental losses following the surface broadcast of P at the NTCC site, a potential alternative to reduce P losses in tile drainage would be the subsurface placement of fertilizer with seed. Indeed, the application of P in the subsurface has been shown to considerably reduce the subsurface leaching of P (Grant et al., 2019a, 2019b) and P losses in tile drainage (Williams et al., 2018).

3.3 Nutrient balances over the study period and relationships with nutrient losses in tile drainage

Although it was not an objective of the current study, data were collected throughout the study period on soil fertility and crop yields for comparison to tile drain losses. Soil at the two farms had similar fertility levels (Table S4), and this remained true throughout the 4-year observation period. Crop yields were also comparable at both sites over the study period. The 2018–2019 crop year had somewhat depressed yields on both sites due to weather impacts. Nutrient additions and crop nutrient balances at both sites were in close alignment with provincial crop production recommendations (Table 1), although at the CT site, P fertilizer was applied to meet crop removal expectations, whereas P application at the NTCC was closer in aligning P fertilizer applications to match provincial soil P test recommendations. The CT site, therefore, received more P fertilizer additions, in part to match P applications to crop removal rates, but also as a consequence of the frequent changes in cropping that occurred due to weather conditions. It is important to note that if cropping had gone according to the plan on these sites (i.e., growers were able to follow their standard 4-year rotation and expected crop yields were achieved), both field sites would have slightly exceeded soil test P fertilizer recommendations but not exceeded crop P removal rates (Table 1). The greater P added at the CT site over the study site was not sufficient to raise the soil fertility to a point where it could be detected beyond the natural variability within the field. It is possible that some of the additional P added over the study period at the CT site may also have been lost in surface runoff, which was not measured in the current study; however, field observations during peak tile flow periods found surface water yields from both sites to be minimal.

TABLE 1. Detailed comparison of four crop-year and water-year balances (2017–2021) and nutrients lost via tile drainage water at the two sites (conventional conservation-till [CT], no-till cover crop [NTCC]).

										Nutrient balances (kg ha^−1)			Nutrient losses in tile drainage (kg ha^−1)
Site	Crop and water year	Crop grown	Yield (Mg/ha)	N removed by crop (kg ha^−1)	P removed by crop (kg ha^−1)	P soil Test Fert. Rec. (kg ha^−1)	N added as fertilizer (kg/ha)	N fixed by crop (kg ha^−1)	P added as fertilizer (kg ha^−1)	N	P (soil test)	P (crop removal)	NO3-N	DRP	TP
CT	2017–2018	Soybeans	3.71	239	22.7	0	12	227	25.3	0	25.3	2.6	9.0	0.418	2.215
	2018–2019	cc/Soybeans	3.3	213	20.1	0	15	198	31.9	0	31.9	11.8	18.4	0.895	3.209
	2019–2020	Soybeans	3.84	248	23.6	0	15	233	31.9	0	31.9	8.3	16.6	0.183	2.104
	2020–2021	Winter wheat (red clover cc)	7.41	148	31.4	8.7	150	0	31.9	2	23.2	0.5	1.0	0.584	2.022
	Four-year nutrient balance—This study									2	112.3	23.2	45	2.080	9.550
	Hypothetical 4-year nutrient balance—Planned rotation									40	57.2	−33.2
NTCC	2017–2018	cc/Soybeans	4.04	261	24.4	0	0	261	0	0	0	−24.4	6.2	0.604	1.552
	2018–2019	cc/Corn	10.95	161	35.8	8.7	213	0	25.3	52	16.6	−10.5	6.1	2.431	3.454
	2019–2020	Soybeans	3.84	248	23.6	0	0	248	0	0	0	−23.6	15.4	0.397	1.271
	2020-21	cc/Soybeans	3.64	235	22.3	0	0	235	0	0	0	−22.3	3.6	0.286	1.043
	Four-year nutrient balance—This study									52	16.6	−80.8	31.3	3.718	7.320
	Hypothetical 4-year nutrient balance—Planned rotation									115	33.2	−57.2

• Note: A hypothetical nutrient balance based on the long-term intended/planned 4-year rotation (corn-soy-soy-winter wheat) for each site is also provided. A crop year spans from the day after harvest of the preceding crop to the harvest date of the current year crop. A water year spans from October 1 to September 30. While these two time periods do not directly overlap, they are quite similar in any year. Crop nutrient removal numbers were sourced from OMAFRA's AgriSuite Nutrient Management Planning decision support tool and are estimated based on the crop yield. The soil P fertilization recommendation based on field-measured Olsen soil test levels of phosphorus (P) in the top 15 cm of topsoil as sourced from OMAFRA's Agronomy Guide for Field Crops, Publication 811 and accessed through OMAFRA's AgriSuite Nutrient Management Planning decision support tool https://www.ontario.ca/page/agrisuite. N fixed is an estimate that assumes soybeans fix only the additional N they need above what is provided through fertilization. In this table: N fixed = N removed by crop—N added as fertilizer. Negative nutrient balances represent a nutrient shortfall from what was removed by the crop, whereas positive balances indicate more nutrient was applied than was removed by the crop. Environmental losses are not explicitly considered in the balance calculations.
• Abbreviations: DRP, dissolved reactive P; TP, total P.

Despite the similarities in soil fertility levels, and the fact that more P fertilizer application events occurred on the CT site, DRP losses were consistently greater in tile drainage from the NTCC site where fertilizer was surface broadcast, even during periods when fertilizer had not recently been applied. Soil fertility levels measured in the 0–15 cm soil profile were slightly higher at the NTCC site (22 vs. 16), and soil P analyses of the top 2 cm of the soil profile suggest a higher degree of soil P stratification at the NTCC site, which may have supplied P into tile drains via macropores. Indeed, saturation of P sorption sites in the top 0–2 cm may have allowed DRP to be more mobile at the NTCC site. In contrast, fertilizer was incorporated at the CT site and therefore more evenly distributed and more likely to become bound to soil particles.

The annual DRP and TP losses in tile drainage at the study sites were greater than what has been observed in other studies within the Ontario Great Lakes region (e.g., Lam et al., 2016; van Esbroeck et al., 2016; Plach et al., 2019), despite the fact that the other Ontario study sites were similar in soil test P levels. Such differences may be a consequence of greater preferential flow between surface soils and tile drains in the finer-textured soil type at the study sites (e.g., Pluer et al., 2020; Macrae et al., 2019), a smaller interaction between subsurface flow and the soil matrix (Grant et al., 2019), and greater geochemical soil P buffering in the subsurface (Plach et al., 2018). Thus, although P losses in tile drainage have been related to soil test P (Duncan et al., 2017; Pease et al., 2018; Plach et al., 2018), these relationships may differ across regions and soil textures.

The fact that edge-of-field P losses in tile drainage are relatively large at these sites despite the fact that crop removal balances are only slightly positive (CT, +23.2) to even negative (NTCC, −80.8) over the 4-year study period suggests that minimizing P application rate and maintaining low soil test values are not the only factors that must be addressed to mitigate environmental P losses. Application timing and fertilizer placement during application are also important variables, although application timing is more difficult to control given the unpredictable nature of weather. Given the incidental losses observed in the current study, avoiding P applications at general times of the year such as the non-growing season when hydrology is more active is advantageous. The significant incidental losses from the surface broadcast P at the NTCC site, despite the presence of a living cover crop at the time of application, suggest that placing the applied fertilizer below ground may be more beneficial.

The considerable PP losses in tile drainage at the CT demonstrate the fact that particulate losses in tile drainage can be significant in finer-textured soils, and that the use of cover crops and no-till (NTCC) practices can considerably reduce these losses. This is relevant to the impairment of surface waters as PP fractions can also become bioavailable (Baker et al., 2014). However, this study also demonstrates that cover crops and no-till alone will also not achieve water quality targets due to the trade-offs associated with elevated DRP losses in tile drainage. This study demonstrates that full attention to all of the “R”s in a “4R” nutrient management strategy is critical if significant water quality improvements are going to be achieved.

Annual NO₃⁻ losses from the field through the tile drainage system were more pronounced on the CT site compared to the NTCC site, despite the fact that both sites had similar amounts of N fertilizer applied over the 4 years of observation. The greatest losses on the NTCC site occurred in the single year following a high N application rate to the corn crop. Although the N fertilizer application in the spring of 2019 was based on a greater anticipated corn yield, weather conditions over the growing season were not conducive to the anticipated yields. As a result, less N was used by the crop than was applied. This, combined with an inability in the late fall of that year to establish a cover crop following corn harvest, likely led to the higher losses of NO₃⁻ to the tile drainage system in the following growing season. The CT site consistently showed higher NO₃⁻ losses in the soybean production years than the NTCC site, suggesting a positive benefit of the cover crop system at the NTCC site.

4 CONCLUSIONS

As farmers and watershed managers in the Great Lakes Region grapple with requirements to meet water quality targets and, in particular the 40% TP and DRP load reductions to Lake Erie, trade-offs between nutrient fractions (dissolved vs. particulate and P vs. N) under different land management practices are increasingly under scrutiny. This study addresses a key knowledge gap about potential trade-offs in nutrient losses from tile drainage, arising from no-till cover crop systems versus more conventional yet conservation-minded tillage practices. The study has shown that tillage increased PP and TSS losses in tile drainage. However, crucially, the smaller PP losses at the NTCC site were offset by greater DRP losses (largely as incidental losses following a single P application), such that TP losses were only slightly greater from the more conventional conservation till (CT) site. Although cover crops and no-till reduced soil erosion and PP and possibly the NO₃⁻ losses at the NTCC site, these alone will not achieve the water quality targets due to the trade-offs from increased DRP losses in tile drainage. Other “4R” strategies such as subsurface placement, along with improved timing of fertilizer applications to avoid hydrologically active periods during the non-growing season, offer opportunities to reduce DRP losses in tile drainage from no-till cropping systems. Although more fertilizer was applied at the CT site over the study period, this did not translate to larger crop yields or greater DRP losses in tile drainage, although larger NO₃⁻ losses occurred. In terms of nutrient management to reduce tile-drain P losses, it is unlikely that reductions in fertilizer P application would be sustainable, as soils at both sites were already at reasonable agronomic levels. A reduction in soil P has become a panacea for improving water quality, but the results of the current study indicate that if soil P levels are not already at high or non-responsive levels, a reduction in soil P alone will be insufficient to achieve water quality targets in tile drainage. This study also demonstrates that, to evaluate the efficacy of management practices and understand potential trade-offs, comprehensive, multi-year sampling programs are needed. Moreover, farmers may have to optimize management practices in response to water quality targets and priority nutrient fractions if these management practices do not impact crop yield.

AUTHOR CONTRIBUTIONS

M. Macrae: Conceptualization; data curation; formal analysis; investigation; methodology; visualization; writing—original draft; writing—review and editing. J. M. Plach: Data curation; methodology; writing—review and editing. R. Carlow: Data curation; investigation. C. Little: Conceptualization; funding acquisition; writing—review and editing. H. P. Jarvie: Writing—review and editing. K. McKague: Conceptualization; formal analysis; writing—review and editing. W. T. Pluer: Visualization; writing—review and editing. P. Joosse: Data curation; writing—review and editing.

ACKNOWLEDGMENTS

The Lower Thames Valley Conservation Authority, especially Dan Bittman, is acknowledged for significant logistical support. Project funding was obtained from the Ontario Ministry of Agriculture, Food and Rural Affairs (OMAFRA) (GLASI program), the Ontario Soil and Crop Improvement Association (ONFARM program; Canadian Agricultural Partnership), and Environment and Climate Change Canada (ECCC). Agriculture and Agrifood Canada (AAFC), especially Natalie Feisthauer, is thanked for providing soil sampling and analysis under Science and Technology Branch project funding. The views expressed in this publication are those of the authors and do not necessarily reflect those of the province of Ontario or the Government of Canada. Two anonymous landowners are gratefully acknowledged for the contribution of their fields and management information.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.